Shake-table test on a historical masonry aggregate: prediction and postdiction using an equivalent-frame model

Modeling the seismic response of historical masonry buildings is challenging due to many aleatory and epistemic uncertainties. Additionally, the interaction between structural units further complicates predictions of the seismic behavior of unreinforced masonry aggregates found throughout European city centers. This motivated the experimental campaign on half-scale, double-leaf stone masonry aggregates within the SERA (Seismology and Earthquake Engineering Research Infrastructure Alliance for Europe)-Adjacent Interacting Masonry Structures project. The experimental campaign included a blind prediction competition that provided participants with data on materials, geometry, construction details, and seismic input. After the test, the actual seismic input and all recorded and processed data on accelerations, base-shear, and displacements were shared with participants. Instead of a single analysis for the prediction phase, we performed broader stochastic incremental dynamic analyses to answer whether the common assumptions for aggregate modeling of either fully coupled or completely separated units yield safe predictions of aggregate behavior. We modeled buildings as equivalent frames in OpenSEES using a newly developed macroelement, which captures both in-plane and out-of-plane failure modes. To simulate the interaction between two units, we implemented a new material model and applied it to zero-length elements connecting the units. Our results demonstrate the importance of explicitly modeling the non-linear connection between the units and using probabilistic approaches when evaluating the aggregate response. Although modeling simplifications of the unit interaction and deterministic approaches might produce conservative results in predicted failure peak ground acceleration, we found that these simplified approaches overlook the likely damage and failure modes. Our results further stress the importance of calibrating material parameters with results from equivalent quasi-static cyclic tests and using appropriate damping models.